Download HW2: Push-Pull Simulation and more Exercises Production and Operations Management in PDF only on Docsity!
Full Name: Tuan Tu Luc – Student ID: 7927453 Full Name: Thi Huong Giang Cao – Student ID: 7959979 Assignment Description for Pull-Push Simulation A key issue in Production Planning and Inventory Control is the choice of mechanism that triggers production in any process or at any workstation (an area where a particular task or set of tasks is done). I.e., if I’m responsible for a workstation in an operating process, how do I know specifically what to do? For example, in a manufacturing process that makes components for three models of computers (laptop, desktop, tablet), and that process requires some changeover time to switch production from one model to another when I finish a batch of one model, how do I know what to do next? The answer to this question has profound implications for the ability to meet demand, inventories, and for capacity utilization. There are two basic mechanisms that are common in practice – PUSH and PULL. A PUSH system relies on a detailed schedule, typically prepared well in advance by Production Planning staff, for each workstation that specifies what is to be produced at each time of the day. For the example above, a very simple PUSH system would say, “Make a batch of x components for the laptop model, then change over to y components for the desktop model, and then when that batch is complete, change over to, and make a batch of size z for the tablet. Repeat this cycle for the day.” A PULL system, in contrast, is reactive, typically to the workstation, which immediately follows it. E.g., if, in the production of these computer components, task A is followed by task B, workstation A will work on whatever model is needed most by Task B. A typical PULL model would thus dictate, “Whenever you finish a batch, ‘look’ downstream and see what task B is running out of, and then change over to, and make a batch [specified size] of the model that B has least of. If B has enough of everything [specify what’s ‘enough’], then stop production” Assignment Objective: · To give students some intuition about production flow in a complex, multi-stage process, and how that flow is influenced by bottleneck and production triggers; · To contrast the dynamics of PUSH, PULL, and mixed systems; · To explore the relations between batch size and set-up times in those systems. Simulation Description This is in some ways a complex process, but it has numerous simplifications compared to many real-world processes, for example, we ignore any transport time between workstations. Every workstation follows a strict batch discipline – batches only move when the entire batch is complete. Each workstation has a batch size, a change-over-time (denoted “set-up” time), and a run time, and is either set to PUSH or PULL. The user can change any of these parameters for any of the workstations. For simplicity, there is a universal settings box at the top of the screen that the user can employ to make changes to all workstations at once. Also notice that the user can change “Run time per unit”, but that
it can only be changed by using this universal settings box. It is pre-loaded with a value of 4 minutes for all scenarios. The parallel workstations that feed a single downstream workstation are meant to be working on components or subassemblies that must be assembled at the next workstation before it begins its work. Thus, for example, workstations G and H must both finish their work on a batch of the same model before both batches can be delivered to L. They are immediately combined at L and go into L’s “raw material” inventory (the quantity of which is shown by the colored bars to the left of each workstation) and become part of the system’s work-in-process inventory. In the same way, the subassemblies at O, P, and Q are combined (instantaneous, in this simplification) to become finished goods at the retailer. Think, for example, of Dell, which combines multiple components and subassemblies to produce various computer models. Using the above example, as an illustration, we have tried to make this process more concrete by picturing the finished product as one of three models of computer. However, in this exercise, the customer arrives and expects the product to be on the retail shelf (e.g, Staples). Any product demanded by a customer is considered a “stockout” if it is not in finished goods at the retailer when the customer arrives. Customer arrivals follow a Poisson process with a mean inter-arrival time of ten minutes. The customer chooses a model at random, with all three models being equally likely. This variation in demand patterns causes some fluctuation in the performance measure results from run to run. The demand process has autocorrelation built in, to reduce the variation in mix. In particular, the same model can’t be demanded twice in a row, which confines the variation to a level that shouldn’t interfere with student learning. Work-In-Process (WIP) inventory cost is $30 per unit (an obvious simplification, since it would build from far less than that for simple electronic raw materials to far more than that for complex subassemblies). Finished goods inventory, at the retailer, is valued at $300 per unit. The exercise runs ten hours per day, seven days per week, for a week. This results in an average demand of (4200 minutes/10 minutes) 420 units. That is a long enough run to dampen the worst variation. See the “How-To-Play Instructions” for more detail on the specifics of the model and the exercise setup. QUESTIONS: Question 1
1. Alter Scenario 2, so that workstation I has a 25-minute set-up time (all others with 20-minute set-up times and batches of 5 units for each of the three models, and all workstations set to PUSH). Before you run it, a. Where is the bottleneck?
Throughput of each workstation O and Q: 10 ∗ 60 20 ∗ 5 = (^150) (units) Throughput of workstation I: 10 ∗ 60 25 ∗ 5 = (^120) (units) Since workstation I pushes 120 units to workstations M and P, the throughput of workstation P is 120. In total, the throughput of the entire production system: (^150) + 150 + 120 = 420 (units) Therefore, in this scenario, the process will be able to keep up with the demand at the end of the day. d. Where, if anywhere, will work-in-process inventory (WIP, shown as a workstation’s “raw material”) build up? The WIP will build up at workstation I due to the longer set-up time. The bottleneck created by the adjusted setup time will result in an increase in WIP inventory as it cannot keep up with the processing pace of other workstations, thus leading to queuing and decreasing the number of batches moving downstream. e. The cost of WIP shown in the performance results seem to be far too high given the WIP that you can see. Where is this “hidden” WIP? The hidden WIP inventory is likely to pile up at workstation I. In fact, although 150 units will be continuously pushed from workstation C to workstation I, workstation I will only have the capacity to process 120 units because of the longer setup time of 25 minutes, resulting in long queues between workstations and an accumulation of hidden WIP, as well as an increase in inventory costs. Question 2 Now click on the “push” arrow on the upper right to change all workstations to PULL. Answer questions (a) – (c) as in question #1 and follow the same directions. What is different? Why? a. Where is the bottleneck? The bottleneck in this scenario will be station I with the longest setup time (25 minutes). In the PULL mode, all downstream workstations will also have to wait for the upstream workstations to complete processing the batches before they can begin processing.
Therefore, the bottleneck will be the workstation with the longest setup time compared to others (20 minutes). b. At which workstation(s) do you expect the lowest capacity utilization? Do you expect the average capacity utilization to be (a) higher than in question 1; (b) lower than in question 1; (c) the same as in question 1; or (d) need more information to be able to answer. Except for workstation I, all other stations have the lowest utilization at 79.9%. The average capacity utilization is expected to be lower than in Question 1. Since the process is now being run in PULL mode, it will lead to longer processing times and more idle time, resulting in a decrease in overall process efficiency. Besides, in PULL configuration, the system operates according to the need of downstream workstations or customer demand, so the upstream workstations only work if the downstream workstations produce less than a full batch of some model, or else, they will stay idle to avoid overproduction. c. Will the process be able to keep up with demand? Yes, the process will be able to keep up with the demand since workstations O, P, and Q are capable of producing 420 units. The calculation is the same as (c) in Question 1. d. Run the exercise again, two or three times. What changes? Why? By running the scenario three times, it can be noticed that there is a slight increase in the utilization of all workstations and the average capacity utilization of the system (80.5% to 80.9%). In addition, the total inventory costs, as well as the inventory in retailers witness a slight fluctuation. Also, the demand for desktops is higher in the third run (144) compared to the first and second runs (139). Since the PULL mode system produces according to the constant customer demand of 420 units, the results may vary slightly due to the random nature of the simulation and variability in customer demand (e.g. more customers demanded desktops, so the factory focused on producing them compared to laptops and tablets in the third run). Questions 3 and 4
3. Go back to the values of question 1 (all PUSH). If you could change one workstation to PULL, which would it be to reduce total inventory cost without incurring stock-outs? Simulate to test out various options. (From here on, if you like, you can choose “Run to End” rather than “Simulate”) Which is best? Why? Workstation O, P, or Q is the ideal workstation to switch to PULL in order to lower total inventory costs without experiencing stock-outs as it will produce a low level of FGI while maintaining a sufficient level of WIP inventory to cope with customer demand. To be more specific, whereas the other upstream workstations (e.g. C, I, and M) are set to PUSH, they will be stimulated to produce as much as possible as quickly as possible.
6. Now, run the process as all PULL. What changes? What doesn’t change? Why? With the PULL setup, the Finished Goods and Work in Process Inventories have significantly reduced, there were more stockouts, the capacity of most workstations did not reach 100% and overall capacity utilization was lower than that in the PUSH setup (80.8%). The unchanged is that the bottleneck is still workstation I. The reasons are that in the PULL setup, the workstations will be monitored by the real demand of the downstream workstations or customer demand, which reduces inventory costs and utilization of workstations to prevent overproduction. 7. Without changing any of the process times, what could you do to improve the process? Run the exercise to test out various alternatives. What’s the best? I would suggest increasing the batch sizes at workstation I, it will help to reduce the backlog in the station without compromising the revenue (if we reduce the batch size, it will reduce the WIP, but the process cannot meet the demand and cause stockouts, leading to lost revenue). **Questions 8 and 9
- Now, make the system run significantly faster. In the upper right corner of the screen, there is an input box to change all set-up times. Change them all to 15 (minutes). Then go to workstation I, and change its set-up time to 20 minutes. Change all batch sizes back to 5 units. This process now has every workstation setting up five minutes faster than the process that you started with in question 1. Run this process in both all-PUSH and all-PULL configurations. What do you see? Why?** PUSH mode: Compared to the previous PUSH mode, the system has experienced a significantly higher total inventory cost, higher utilization of each workstation (most stations reach a utilization of 100%, while workstations M and P have utilization of 87.5%), and higher overall capacity utilization (98.5%), no stockouts and more inventories in retailers. In addition, the system fails to fulfill the demand of customers since it could only produce 419 units. PULL mode: Compared to the previous PULL mode, the system has witnessed an increase in total inventory cost with a difference of around $1,915, lower utilization of each workstation and overall capacity utilization (70.8%), and no stockouts and more inventories in retailers. Besides, the demand from the customer is met. In general, however, the bottleneck is still workstation I as it remains the slowest station in the entire production process and has the highest utilization in both settings. To explain the differences, reducing the setup time at all stations will decrease the total cycle time of the production process, resulting in higher total throughput to meet the customer demand. Additionally, in the PUSH setting, production is operated based on estimated demand by scheduling the release rate of each workstation. In the PULL setting, however, orders are fulfilled in response to real market demand or the needs of downstream workstations. Hence, utilizations of all workstations and capacity utilization are higher in the PUSH setting than in the PULL setting.
9. Without changing any of the process times, what could you do to minimize batch sizes, while still keeping stockouts minimal? Run the exercise to test out various alternatives. What’s the best? Is all PULL better than all PUSH? Along what dimensions? To minimize the batch size, I propose reducing the batch size at each workstation. And the optimal batch size should be 4. This will foster the system to better fulfill the demand and minimize stockouts at 1, with capacity utilization of around 78%. However, we cannot reduce the batch size to less than 4 since it makes the production system fail to meet the demand and significantly inflate stockouts. For example, when setting the batch to 3, the system can only produce 393 units compared to 420 units of customer demand, and stockouts incur at 350. In addition, when the batch size is decreased to 2, only 300 units are produced and stockouts are much worse at 413. All PULL settings are better than PUSH settings. The PULL configuration contributes to lowering the cycle time and inventory levels, minimizing inventory costs. The PUSH configuration is better at reducing stockouts.
